Is the space $\psi$ (described in problem 5I of L. Gillman and M. Jerison, Rings of continuous functions, Springer Verlag, 1976) a F-Z-space (i.e, space with $cl(X-Z(f))$ is a zero set for every $f$ in $C(X)$)?

$\textbf{Clarification}$ A collection $\mathcal{A}$ of infinite subsets of $\mathbb{N}$ is said to be an almost disjoint family if $R\cap S$ is finite whenever $R,S\in\mathcal{A}$ and $R\neq S$.

Let $\mathcal{E}$ be a maximal almost disjoint family of subsets of $\mathbb{N}$. Let $D=(w_{E})_{E\in\mathcal{E}}$ be a set of new distinct points. Let $\psi=\mathbb{N}\cup D$. We shall give $\psi$ the topology as follows. The points in $\mathbb{N}$ are isolated. The neighborhood filter around a point $w_{E}$ is generated by sets of the form $w_{E}\cup R$ where $R\subseteq E$ and $E\setminus R$ is a finite set. In this space, $w_{E}\cup E$ is the one-point-compactification of the space $E$.

The problem in the book states that the space $\psi$ is completely regular, pseudocompact, but not countably compact. In particular, $\psi$ is not realcompact and $\psi$ is not normal. Furthermore, every subspace of $\psi$ is a $G_{\delta}$-set.

  • 7
    $\begingroup$ Can you please include the definitions of each thing that appears in this question? At least, you should take the time to define \psi $\endgroup$
    – Jon Bannon
    Commented Sep 2, 2013 at 16:28
  • 1
    $\begingroup$ If you are trying to do Problem 5I, please give some details of where you are stuck $\endgroup$
    – Yemon Choi
    Commented Sep 2, 2013 at 19:02
  • 2
    $\begingroup$ I checked problem 5I and I found nothing about F-Z-spaces in that problem. Also, I edited this question to include the definition of $\psi$ and information about this space to clear any confusion. I am voting to reopen this question. $\endgroup$ Commented Sep 3, 2013 at 3:06
  • 1
    $\begingroup$ For a normal but not FZ-space, start with a Tychonoff plank and add a convergent sequence $\langle s_m^n:n<\omega\rangle$ to each point $[\omega_1, m]$ on the right-most vertical column. Then the set $S=\{s_m^n:n,m<\omega\}$ is a co-zero set whose closure is not $G_\delta$ hence not zero. $\endgroup$ Commented Sep 3, 2013 at 10:15
  • 1
    $\begingroup$ Another way to construct examples of normal spaces which are even compact and connected but are not FZ-spaces is to look at the order topology. For instance, if one takes the long line and adds a copy of the unit interval $[0,1]$ at the end of the long line, one obtains a ordered space $([0,1)\times\omega_{1})\cup[0,1]$ which is compact and connected since it is a complete lattice. The interval $(0,1]$ on the very right is a cozero set, but the closure $[0,1]=\overline{(0,1]}$ is not a zero set in $([0,1)\times\omega_{1})\cup[0,1]$. $\endgroup$ Commented Sep 3, 2013 at 14:53

1 Answer 1


As I commented above, I think the answer is that a $\psi$-space need not be an FZ-space, and that a counterexample may be constructed from a Luzin gap. Here are the details, which did not fit into the comment.

We first construct the MAD family, which will give the counterexample. Start by splitting ${\mathbb N}$ into two infinite sets $X_0,X_1$. Then let ${\mathcal A}_0=\{A^0_\alpha:\alpha<\omega_1\}$ be an arbitrary AD family on $X_0$ and ${\mathcal A_1}=\{A^1_\alpha:\alpha<\omega_1\}$ a Luzin gap on $X_1$ (i.e. an AD family which, in particular, has the property that no uncountable pair of subfamilies can be separated; here ${\mathcal B},{\mathcal C}\subseteq{\mathcal A}_1$ can be separated if there is a set $S\subseteq X_1$ which almost-contains all sets from $B$ and is almost disjoint from all sets in $C$.).

We shall fuse these two families into a single new AD family which we will extend to a MAD family. Proceed as follows: for $\alpha<\omega_1$ let $A_\alpha=A^0_\alpha\cup A^1_{\alpha+1}$ and let $\{B_\alpha:\alpha<\omega_1\}$ be the enumeration of $\{A^1_\beta:\beta\in Lim(\omega_1)\}$. Let ${\mathcal A}$ be an arbitrary MAD family extending the family $\{A_\alpha,B_\alpha:\alpha<\omega_1\}$.

Proposition The space $\psi({\mathcal A})$ is not an FZ-space.

Note In the following we identify the points of $\psi({\mathcal A})={\mathbb N}\cup D$ which are in $D$ with the elements of the MAD family ${\mathcal A}$.

Proof. Define $f:\psi({\mathcal A})\to{\mathbb R}$ as follows: $f(n)=1/n$ for $n\in X_0$ and $f(x)=0$ otherwise. It is clear that $f$ is continuous. Let $G=\psi(\mathcal{A})\setminus f^{-1}(0) = X_0$. To show that $\psi({\mathcal A})$ is not an FZ-space it will be sufficient to show that $F=cl(G)$ is not a zero set. Notice first that for each $\alpha<\omega_1$ we have $A_\alpha\in F$ and $B_\alpha\not\in F$. Suppose now that $g:\psi({\mathcal A})\to{\mathbb R}$ is a continuous function such that $g^{-1}(0)=F$. Since $B_\alpha\not\in F$ for each $\alpha<\omega_1$, we can fix $n<\omega$ such that ${\mathcal B}=\{B_\alpha:g(B_\alpha)>1/n\}\subseteq{\mathcal A}_1$ is uncountable. Now let $S=\{k<\omega:g(k)>1/n\}$. Then $S$ almost contains all sets from ${\mathcal B}$ while, by continuity of $g$, being almost disjoint from every set in $\{A_\alpha:\alpha<\omega_1\}$ hence also from every set in ${\mathcal C}=\{A^1_{\alpha+1}:\alpha<\omega_1\}\subseteq{\mathcal A}_1$. This contradicts the fact that ${\mathcal A}_1$ is a Luzin gap, since $S$ separates the two uncountable subfamilies ${\mathcal B,C}\subseteq{\mathcal A}_1$. $\square$

  • $\begingroup$ excuse me. I edit "FZ-space"----> "SZ-space" $\endgroup$ Commented Sep 7, 2013 at 17:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.